This paper goes beyond traditional political budget cycles studies by considering the impact of the election calendar on the composition of tax revenue (direct taxes versus indirect taxes) rather than on the global level. We develop a theoretical model, based on Drazen and Eslava (2010) to predict how the taxation structure will be modif ied during election years. Using a panel of 56 developing countries over 1980-2006, our study reveals clear patterns of electorally timed interventions. We found robust evidence that indirect taxes decreases are the preferred vehicle for incumbents in de veloping countries to increase their popularity just before elections. On average, they are falling of 2.6 percent in an election year while the direct taxes remain unchanged. These manipulations constitute reversals in the developing countries' tax reforms aim- ing at broaden tax bases and increase tax mobilization and point at the importance of both good fiscal institutions and fiscal discipline.